Re: [PATCH 2/2] Enable v4 mounts when either "nfsvers=4" or "vers=4" option are set (vers-02)

From: Chuck Lever <chuck.lever@oracle.com>
To: Steve Dickson <SteveD@redhat.com>
Cc: Linux NFS Mailing list <linux-nfs@vger.kernel.org>,
	Linux NFSv4 mailing list <nfsv4@linux-nfs.org>
Subject: Re: [PATCH 2/2] Enable v4 mounts when either "nfsvers=4" or "vers=4" option are set (vers-02)
Date: Wed, 26 Aug 2009 15:50:21 -0400	[thread overview]
Message-ID: <B26EBCDF-5836-440E-81D5-5E089158BFE1@oracle.com> (raw)
In-Reply-To: <4A95760C.9000604@RedHat.com>

On Aug 26, 2009, at 1:51 PM, Steve Dickson wrote:
> On 08/26/2009 01:22 PM, Chuck Lever wrote:
>>>> and if your proposed method to handle -t nfs -o vers=4 will make
>>>> it more complicated to get there.
>>> No. I'm proposing a simple shorthand patch that will make mounting  
>>> nfs4
>>> file systems easier in hope of moving the technology forward by  
>>> making
>>> it more accessible... What I believe you are proposing is  
>>> architecture
>>> change to hide the fact nfs and nfs4 are separate file systems...
>>
>> Nope, we're proposing doing the simple method in the kernel instead  
>> of
>> in the mount command.
>>
> My apologize then... I was misinterpreting what you guys were  
> suggesting..
> (email sometimes causes that... :-\ )
>
> I don't think the -o v4 translation will be as easy as a
> "simple method in the kernel" and it surely will not be as simple
> and unintrusive as the patch I'm proposing.... Here is why...
>
> From an Linux  architecture standpoint the mount command *always*
> know what file system its mounting. There not been a precedence set
> where mount, mounts one file system which turns into mounting a
> different file system. Meaning there is no kernel support for
> nfs_get_sb() to all of sudden decide to roll back the system call
> and jump into nfs4_get_sb() (or vice a versa depending on which is the
> default).

Yeah, switching file system types in the mount(2) system call is the  
fly in the ointment.  I'm just wondering if Trond had some thoughts  
about making that more feasible.

> Of course we could set that precedence and quit frank we would have
> to. I'm not totally against that, although other in the kernel  
> community
> might be...
>
> But there is no way that re-architecturing of kernel will as simple
> and straightforward as following the Linux standard architecture of
> having mount, mount the correct file system.

--
Chuck Lever
chuck[dot]lever[at]oracle[dot]com